delonix system
O motor em si: events, info, df, prune, monitor, thermal.
The engine itself: events, info, df, prune, monitor, thermal.
Introspecção e manutenção. system prune é o GC (recupera espaço: containers
parados, dirs órfãos, imagens dangling, blobs CAS, hostfwds órfãos, redes vazias); system df
mostra o uso de disco; system monitor segue ligações/conntrack; system events o
fluxo de eventos.
Introspection and maintenance. system prune is the GC (reclaims
space: stopped containers, orphan dirs, dangling images, CAS blobs, orphan hostfwds, empty
networks); system df shows disk usage; system monitor follows
connections/conntrack; system events the event stream.
📄 Implementação real em Rust: cmd/system.rs
Usage: delonix system [OPTIONS] <COMMAND>
Commands:
df Disk usage by area (images, containers, volumes, VM images)
doctor Check the HOST prerequisites this engine needs, and say how to fix each
events Engine events (create/start/die/remove/…), from oldest to most recent
features What each capability promises, and the EVIDENCE for it
info Engine state: rootless?, cgroup delegation, network infra, counts
metrics The raw counters: containers/VMs/networks/volumes/images, memory, network, disk
namespace Isolation namespaces: ls/describe — a namespace exists while something is in it, so there is no create/rm
resources How much this host has, how much the engine can actually enforce, and what is under pressure right now
regulate Give the contended CPU back to the workloads that are waiting for it
setup Diagnose — and, with `--delegate`, fix — cgroup delegation
snapshot Point-in-time capture of this node's whole state — `create` saves it into one archive, `restore` puts it back
virt Host virtualization: hypervisor, KVM, virtio — and what there is to tune
prune Reclaim space taken by what nothing uses any more
thermal Thermal governor for Delonix's CPU budget. Runs continuously (see `--once`)
monitor Active network connections per container (via conntrack)
boot Boot persistence: systemd units so containers come back up after a reboot
help Print this message or the help of the given subcommand(s)
Options:
--l18n <en|pt>
Output language: `en` (default) or `pt` (Portuguese, pt_AO). Also settable via `$DELONIX_L18N`. Global — works before any subcommand
-h, --help
Print help (see a summary with '-h')
COMMAND MAP:
Inspect events · info · features · doctor · resources · metrics · df · namespace
Configure setup · regulate · snapshot · virt
Maintenance prune · thermal
Dashboards monitor
Advanced boot
EXAMPLES:
# is this rootless, is the cgroup delegated, what is the network doing
delonix system info
# where the disk went
delonix system df
# get it back, orphan container directories included
delonix system prune -f
SEE ALSO:
delonix dashboard · delonix system events · delonix container stats ·
delonix image ls
delonix › systemsystem thermal
Thermal governor for Delonix's CPU budget. Runs continuously (see --once).
Lowers the budget when the CPU heats up and restores it when it cools down.
Usage: delonix system thermal [OPTIONS]
Options:
--high <HIGH>
Temperature (°C) at or above which it cools down
[default: 85]
--l18n <en|pt>
Output language: `en` (default) or `pt` (Portuguese, pt_AO). Also settable via `$DELONIX_L18N`. Global — works before any subcommand
--low <LOW>
Temperature (°C) below which it restores
[default: 70]
--floor <FLOOR>
Minimum CPU percentage it drops to
[default: 40]
--interval <INTERVAL>
Seconds between readings
[default: 5]
--once
One reading and exit (for cron/scripts, instead of the loop)
-h, --help
Print help (see a summary with '-h')
EXAMPLES:
# keep the engine's CPU budget in check when the machine heats up
delonix system thermal
# one reading and out, for cron
delonix system thermal --once
# cool down earlier, and never drop below half the CPU
delonix system thermal --high 75 --floor 50
SEE ALSO:
delonix system monitor · delonix container update · delonix system info
delonix › system › thermalExemplosExamples
delonix system thermalsystem monitor
Active network connections per container (via conntrack).
Who comes in, who goes out, and between containers. Refreshes continuously (see --no-stream).
Usage: delonix system monitor [OPTIONS]
Options:
--interval <INTERVAL>
Milliseconds between refreshes (minimum 300)
[default: 1000]
--l18n <en|pt>
Output language: `en` (default) or `pt` (Portuguese, pt_AO). Also settable via `$DELONIX_L18N`. Global — works before any subcommand
--no-stream
One sample and exit (without clearing the screen or repeating)
-h, --help
Print help (see a summary with '-h')
EXAMPLES:
# who talks to whom, per container, live from conntrack
delonix system monitor
# one sample and out — the shape for a script or a cron line
delonix system monitor --no-stream
# a slower refresh on a busy host
delonix system monitor --interval 3000
SEE ALSO:
delonix dashboard · delonix network ls · delonix system events · delonix
container stats
delonix › system › monitorExemplosExamples
delonix system monitorsystem virt
Host virtualization: hypervisor, KVM, virtio — and what there is to tune
Usage: delonix system virt [OPTIONS]
Options:
--l18n <en|pt>
Output language: `en` (default) or `pt` (Portuguese, pt_AO). Also settable via `$DELONIX_L18N`. Global — works before any subcommand
--tune
Apply the recommended tuning (needs root)
-h, --help
Print help (see a summary with '-h')
EXAMPLES:
# what the host offers a VM: hypervisor, KVM, virtio — and what is left to
# tune
delonix system virt
# apply the recommended tuning (needs root)
delonix system virt --tune
SEE ALSO:
delonix vm create · delonix get · delonix system info
delonix › system › virtExemplosExamples
delonix system virtsudo delonix system virt --tunesystem setup
Diagnose — and, with --delegate, fix — cgroup delegation.
It is the prerequisite for --memory/--cpus/--pids-limit to mean anything. Without it the flags are ACCEPTED and INERT: the container runs with no limit at all. That is the failure mode this command exists for, and it is invisible without asking.
Usage: delonix system setup [OPTIONS]
Options:
--delegate
Apply the fix instead of only reporting it. Writing the system-wide drop-in needs root; the per-session remedy never does
--l18n <en|pt>
Output language: `en` (default) or `pt` (Portuguese, pt_AO). Also settable via `$DELONIX_L18N`. Global — works before any subcommand
-h, --help
Print help (see a summary with '-h')
EXAMPLES:
# diagnose cgroup delegation — without it --memory and --cpus are accepted
# and inert
delonix system setup
# apply the fix instead of only reporting it (the system-wide drop-in needs
# root)
delonix system setup --delegate
SEE ALSO:
delonix system info · delonix container run · delonix container update ·
delonix cluster create
delonix › system › setupExemplosExamples
delonix system setupsystemd-run --user --scope -p Delegate=yes -- delonix system setupsudo delonix system setup --delegatesystem events
Engine events (create/start/die/remove/…), from oldest to most recent.
With no daemon, the log is a shared append-only file — each command appends its own line (see delonix_node::events).
Usage: delonix system events [OPTIONS]
Options:
-f, --follow
Follow continuously (Ctrl-C to exit)
[alias: -w]
--l18n <en|pt>
Output language: `en` (default) or `pt` (Portuguese, pt_AO). Also settable via `$DELONIX_L18N`. Global — works before any subcommand
-n, --tail <TAIL>
Show only the last N (default: all)
-o, --output <OUTPUT>
Output format: `table` (default) or `json` (ADR-0005). With `-f`, `json` streams ONE OBJECT PER LINE (JSONL) — a JSON array would never close on a stream
[default: table]
[possible values: table, json]
-h, --help
Print help (see a summary with '-h')
EXAMPLES:
# what the engine did, oldest first — with no daemon this is a shared
# append-only log
delonix system events
# only the last twenty lines
delonix system events -n 20
# follow it live while you reproduce a crash in another terminal
delonix system events -f
SEE ALSO:
delonix system info · delonix container logs · delonix system monitor
delonix › system › eventsExemplosExamples
delonix system events -n 20delonix system events --followsystem prune
Reclaim space taken by what nothing uses any more.
Removes stopped containers, unused images, CAS blobs nobody references, empty cgroups and — the biggest space saver — orphan container directories (from nodes/containers that died abruptly without rm, with no registry entry).
Usage: delonix system prune [OPTIONS]
Options:
-f, --force
Skip the confirmation prompt (REQUIRED when stdin is not a terminal)
--l18n <en|pt>
Output language: `en` (default) or `pt` (Portuguese, pt_AO). Also settable via `$DELONIX_L18N`. Global — works before any subcommand
-a, --all
Also remove unused images that DO have a tag (not just the dangling ones)
--auto
SCHEDULED mode: reclaim only when the disk is at or above `--threshold`, and do nothing at all below it.
This is the flag that makes `system prune` safe to put in a timer. Below the threshold it exits 0 having changed nothing, and that is SUCCESS, not a no-op to be fixed: pruning every night by habit deletes images that get pulled again in the morning, which costs bandwidth and start-up time and reclaims nothing that was a problem.
It implies `--force` — an unattended sweep has nobody to answer a prompt — and it never touches volumes (see `volume prune`).
--threshold <THRESHOLD>
Occupancy percentage at or above which `--auto` acts.
Measured the way `df` measures it, on the filesystem holding the engine's state. Keep it BELOW the threshold of whatever alerts on the same disk: the GC has to act before someone is woken up, and with equal numbers the alert always arrives first and the GC looks useless.
[default: 75]
--dry-run
Print what this sweep WOULD take, split by category, and take nothing.
The split is the reason it exists. `prune` mixes two different things in one verb — resources somebody DECLARED (stopped containers, tagged images) and debris that never had a record (orphan directories, unreferenced blobs) — and only the second is safe to hand to a timer. Whoever arms `--auto` has to be able to see which half it would touch first.
Combines with `--auto`: it then also says whether the threshold gate would let the sweep run at all right now.
The three network reapers (orphan slirps, ingress refs, host ports) are NOT previewed — they compute and act in one call inside `delonix-sdn`. None of them frees a byte of disk, so everything that reclaims space is in the report.
-h, --help
Print help (see a summary with '-h')
EXAMPLES:
# reclaim space: stopped containers, unused images, unreferenced blobs,
# orphan rootfs directories
delonix system prune
# in CI, where there is no terminal to confirm at
delonix system prune -f
# from a nightly timer: reclaim only when the disk is at 75% or above, and
# do nothing at all below it
delonix system prune --auto --threshold 75
# also drop tagged images nobody uses, not just the dangling ones
delonix system prune -a -f
# see what it WOULD take, split into declared resources and debris, and take
# nothing — the report to read before putting `--auto` in a timer
delonix system prune --dry-run
# what the scheduled sweep would do right now, threshold gate included
delonix system prune --dry-run --auto
SEE ALSO:
delonix system df · delonix image remove · delonix container rm · delonix
volume ls
delonix › system › pruneExemplosExamples
delonix system prunesystem df
Disk usage by area (images, containers, volumes, VM images)
Usage: delonix system df [OPTIONS]
Options:
--l18n <en|pt>
Output language: `en` (default) or `pt` (Portuguese, pt_AO). Also settable via `$DELONIX_L18N`. Global — works before any subcommand
-h, --help
Print help (see a summary with '-h')
EXAMPLES:
# disk usage by area: images, containers, volumes and VM images
delonix system df
SEE ALSO:
delonix system prune · delonix image ls · delonix volume ls · delonix system
info
delonix › system › dfExemplosExamples
delonix system dfsystem info
Engine state: rootless?, cgroup delegation, network infra, counts
Usage: delonix system info [OPTIONS]
Options:
--l18n <en|pt>
Output language: `en` (default) or `pt` (Portuguese, pt_AO). Also settable via `$DELONIX_L18N`. Global — works before any subcommand
-h, --help
Print help (see a summary with '-h')
EXAMPLES:
# engine state in one screen: rootless, cgroup delegation, network infra,
# counts
delonix system info
# the same screen in Portuguese
delonix system info --l18n pt
SEE ALSO:
delonix system setup · delonix system df · delonix system virt · delonix
dashboard
delonix › system › infoExemplosExamples
delonix system infoLaboratórioLab
Enche disco com lixo (containers parados) e depois recupera-o.
for i in 1 2 3; do delonix container run --name lixo-$i alpine true; done
delonix system df
delonix system prune
delonix system dfFill up disk with junk (stopped containers) and then reclaim it.
for i in 1 2 3; do delonix container run --name junk-$i alpine true; done
delonix system df
delonix system prune
delonix system dfDesafioChallenge
Segue system events num terminal enquanto corres
container run/stop/rm noutro, e mapeia cada acção da CLI ao
evento exacto que ela emite.
Follow system events in one terminal while running
container run/stop/rm in another, and map each CLI action
to the exact event it emits.